Question

Does `clusterCTSS` and `aggregateTagClusters` sum raw-tag counts (and tpm) of the clustered tags into the resultant cluster?

0

Entering edit mode

regcastillo1990 • 0

@regcastillo1990-20939

Last seen 5.5 years ago

I want to keep the original counts as raw as possible (without tpm), even after annotation. In case of clustered tags I need to sum up the counts of the tags merged in to a cluster.

Does clusterCTSS and aggregateTagClusters sum raw-tag counts (and tmp) of the clustered tags into the resultant cluster? Are there a way to force sum of counts, ignoring tmp or other treatment of counts? Which column reflect the sum of counts?

I set method="none", this is my code:

normalizeTagCount(ce, method = "none", fitInRange = c(10, 1000), alpha = 1.25, T = 10^6)

clusterCTSS( ce, threshold = 1, thresholdIsTpm = TRUE, nrPassThreshold = 1, method = "distclu", maxDist = 20, removeSingletons = TRUE, keepSingletonsAbove = 5)

aggregateTagClusters(ce, tpmThreshold = 5, excludeSignalBelowThreshold = TRUE, qLow = NULL, qUp = NULL, maxDist = 100, useMulticore = FALSE, nrCores = NULL)

ce is my CAGEexp object.

CTSS clustering tag per million tag counts CAGEr • 1.5k views

ADD COMMENT • link updated 3.0 years ago by Charles Plessy ▴ 180 • written 5.5 years ago by regcastillo1990 • 0

1

Entering edit mode

clusterCTSS and aggregateTagClusters calculate both the raw tag count and the normalised expression scores of the consensus clusters. It is not possible to skip the calculation of the normalised expression scores (at this point, I am not sure if this would bring a significant performance advantage).

The raw tag counts of each consensus cluster can be found in the counts assay of the the SummarizedExperiment object returned by the consensusClustersSE function. It is a DataFrame of Rle-encoded integers. See for instance the example data:

> head(assay(consensusClustersSE(exampleCAGEexp)))
                 Zf.unfertilized.egg Zf.high Zf.30p.dome Zf.prim6.rep1 Zf.prim6.rep2
chr17:26027430:+                   0       0           0             1             0
chr17:26050540:+                   1       0           0             0             0
chr17:26118088:+                   0       0           0             1             0
chr17:26142853:+                   0       0           0             0             1
chr17:26166954:+                   0       0           0             1             0
chr17:26222417:+                   0       1           1             0             0

ADD REPLY • link 5.5 years ago Charles Plessy ▴ 180

0

Entering edit mode

Thank you Plessy, that solved my present problem.

When I tried CTSStagCountGR(ce) I got this message

Error in samples %in% sampleLabels(object) : argument "samples" is missing, with no default

Where this samples come from? What did I miss?

This is what sampleLabels(ce) return of my ce object

             #FF0000FF              #00FF00FF              #0000FFFF 
"t1.BAM" "t2.BAM" "t3.BAM"

ADD REPLY • link 5.5 years ago regcastillo1990 • 0

1

Entering edit mode

CTSStagCountGR works at the level of individual samples, so it needs a sample argument. If you want all CTSS positions, you can use CTSScoordinatesGR If you want an expression table, you can use CTSStagCountDF.

ADD REPLY • link 5.5 years ago Charles Plessy ▴ 180

0

Entering edit mode

Thank you C. Plessy. I successfully completed a pipeline with CAGEr to get the counts per sample of each consensus clustered tags, with at most 12 libraries. I added other three libraries but got this error.

Error in validObject(.Object) : invalid class “ConsensusClusters” object: Consensus clusters must not overlap with each other

I run a set with only these three libraries, and get a reasonable output, without error, also combined with others. It depends on the combination of libraries I choose, and the error re-reappear. I need to process all libraries together, this time 21, but I couldn't pass from 12, and not all combination of 12 in 21 escape from this error message.

Is it not the job of aggregateTagClusters function, to aggregates tag clusters into non-overlapping consensus clusters? Should I care for parameters or something else, to made it possible?

I use the same parameters of my initial post, above in this thread. any advise?

ADD REPLY • link 5.5 years ago regcastillo1990 • 0

0

Entering edit mode

Hi, if you still have interest in CAGEr, maybe you can try our version 2.0; we fixed a lot of bugs there!

ADD REPLY • link 3.0 years ago Charles Plessy ▴ 180